[Samba] kcc_periodic output

[Douglas Bagnall]

On 9/02/24 06:35, Elias Pereira wrote:
>     for i in $(seq 1 10); do time /usr/sbin/samba_kcc --debug; sleep 5;done
> 
> I ran the command and in one run it lasted more than 2 minutes.
> 
> I've sent you the file if you'd like to take a look.

To summarise for the list, one of the runs took 3 minutes, and the rest of them 
took half a second.

$ grep -A2 real Downloads/kcc_output.txt
real	2m59.646s
user	0m0.387s
sys	0m0.163s
--
real	0m0.504s
user	0m0.413s
sys	0m0.078s
--
real	0m0.496s
user	0m0.371s
sys	0m0.112s
--
[...]

The debug output of each run is identical.

Looking at the 'user' and 'sys' lines, we can see that samba_kcc wasn't doing 
anything extra in those 3 minutes. It was just waiting. For what, I don't know.

Douglas



> 
> On Wed, Feb 7, 2024 at 8:03 PM Douglas Bagnall <douglas.bagnall at catalyst.net.nz 
> <mailto:douglas.bagnall at catalyst.net.nz>> wrote:
> 
>     On 8/02/24 11:17, Elias Pereira wrote:
>      > Hi Douglas, thanks for the answer!!!
>      >
>      >     samba kcc command = /path/to/samba_kcc --debug
>      >
>      >
>      > I ran the command and the result is in the pastebin below, if you can
>      > take a look.
>      > https://pastebin.com/x6W2yWm7 <https://pastebin.com/x6W2yWm7>
>     <https://pastebin.com/x6W2yWm7 <https://pastebin.com/x6W2yWm7>>
> 
>     It looks to me like that was a successful run, but I can't see if it was
>     slow.
> 
>     If I understand correctly, sometimes it is bad, and sometimes it is not.
>     If you run in a loop, something like
> 
>     for i in $(seq 1 10); do time /usr/sbin/samba_kcc --debug; sleep 5;done
> 
>     then you might find one that takes more than 40 seconds -- that is the
>     NT_STATUS_IO_TIMEOUT.
> 
>     Douglas
> 
> 
>      >
>      >     But anyway, I don't think this is a samba_kcc specific problem, and
>      > https://bugzilla.samba.org/show_bug.cgi?id=15568
>     <https://bugzilla.samba.org/show_bug.cgi?id=15568>
>      >     <https://bugzilla.samba.org/show_bug.cgi?id=15568
>     <https://bugzilla.samba.org/show_bug.cgi?id=15568>>is due to the same
>      >     underlying issue. I don't know what that is.
>      >
>      >
>      > I opened this bug, but no one has replied yet.
>      >
>      >     >> UpdateRefs failed with NT_STATUS_IO_TIMEOUT
>      >     This is a different replication failure, independent of the KCC.
>      >     samba_kcc works out where the DC should replicate from, and the drepl
>      >     code does the actual replication.
>      >     Both could fail with network weirdness, or if either the local or remote
>      >     database is locked up for some reason.
>      >
>      >
>      > And how can I debug this problem? I've searched everywhere and found
>      > nothing about it.
>      >
>      > On Wed, Feb 7, 2024 at 6:40 PM Douglas Bagnall
>      > <douglas.bagnall at catalyst.net.nz <mailto:douglas.bagnall at catalyst.net.nz>
>      > <mailto:douglas.bagnall at catalyst.net.nz
>     <mailto:douglas.bagnall at catalyst.net.nz>>> wrote:
>      >
>      >     On 8/02/24 02:11, Elias Pereira via samba wrote:
>      >      > Any idea how I can debug this?
>      >
>      >
>      >      >>> I don't know the answer to that, but I think setting
>      >      >>>
>      >      >>>       samba kcc command = /usr/sbin/samba_kcc -d5
>      >      >>>
>      >
>      >     I looked it up. Actually
>      >
>      >              samba kcc command = /path/to/samba_kcc --debug
>      >
>      >     *might* produce more interesting results (you'll probably want to use
>      >     `less -R` to view the logs).
>      >
>      >     You can run samba_kcc independently -- it should not do any harm, but
>      >     you can use the --readonly flag if you want to be careful.
>      >
>      >     But anyway, I don't think this is a samba_kcc specific problem, and
>      > https://bugzilla.samba.org/show_bug.cgi?id=15568
>     <https://bugzilla.samba.org/show_bug.cgi?id=15568>
>      >     <https://bugzilla.samba.org/show_bug.cgi?id=15568
>     <https://bugzilla.samba.org/show_bug.cgi?id=15568>>
>      >     is due to the same underlying issue. I don't know what that is.
>      >
>      >      >> [2024/02/02 20:05:17.752322,  0]
>      >      >>
>      >   
>       ../../source4/dsdb/repl/drepl_out_helpers.c:1300(dreplsrv_update_refs_done)
>      >      >>    UpdateRefs failed with NT_STATUS_IO_TIMEOUT
>      >
>      >     This is a different replication failure, independent of the KCC.
>      >
>      >     samba_kcc works out where the DC should replicate from, and the drepl
>      >     code does the actual replication.
>      >
>      >     Both could fail with network weirdness, or if either the local or
>      >     remote
>      >     database is locked up for some reason.
>      >
>      >      >> What does this part of the code do?
>      >      >> source4/dsdb/kcc/kcc_periodic.c:790
>      >
>      >     That just reports that samba_kcc failed.
>      >
>      >     cheers,
>      >     Douglas
>      >
>      >
>      >
>      > --
>      > Elias Pereira
> 
> 
> 
> -- 
> Elias Pereira